curioustorvald/tsvm
1
0
Fork 0
mirror of https://github.com/curioustorvald/tsvm.git synced 2026-03-13 22:46:06 +09:00
Code Issues Packages Projects Releases Wiki Activity

ipf codec wip

Browse Source
This commit is contained in:
minjaesong
2022-05-01 15:44:53 +09:00
parent 687cc181e6
commit 37d6f76a5f
2 changed files with 332 additions and 31 deletions
Download Patch File Download Diff File Expand all files Collapse all files

114
assets/disk0/encodeipf.js
View File

@@ -33,6 +33,7 @@ Image is divided into 4x4 blocks and each block is serialised, then the entire f
Bits are packed like so:
uint32 SUBSAMPLING MASK (unimplemented; dont write this)
uint8 [Cg-Top Left | Co-Top Left]
uint16 [Y1 | Y0 | Y5 | Y4]
uint8 [Cg-Top Right | Co-Top Right]
@@ -41,21 +42,38 @@ Image is divided into 4x4 blocks and each block is serialised, then the entire f
uint16 [Y9 | Y8 | YD | YC]
uint8 [Cg-Bottom Right | Co-Bottom Right]
uint16 [YB | YA | YF | YE]
(total: 96 bytes)
(total: 16 bytes)
If has alpha, append following bytes for alpha values
uint16 [a1 | a0 | a5 | a4]
uint16 [a3 | a2 | a7 | a6]
uint16 [a9 | a8 | aD | aC]
uint16 [aB | aA | aF | aE]
(total: 160 bytes)
(total: 24 bytes)
Subsampling mask:
Least significant byte for top-left, most significant for bottom-right
For example, this default pattern
00 00 01 01
00 00 01 01
10 10 11 11
10 10 11 11
turns into:
01010000 -> 0x30
01010000 -> 0x30
11111010 -> 0xFA
11111010 -> 0xFA
which packs into: [ 30 | 30 | FA | FA ] (because little endian)
*/
if (!exec_args[2]) {
printerrln("Usage: entsvmipf input.jpg output.ipf")
printerrln("Usage: encodeipf input.jpg output.ipf")
return 1
}
@@ -74,23 +92,42 @@ dma.comToRam(0, 0, infile, fileLen)
// decode
const [imgw, imgh, imageData, channels] = graphics.decodeImage(infile, fileLen) // stored as [R | G | B | (A)]
sys.free(infile)
let hasAlpha = (4 == channels) && configUseAlpha
let blockCount = Math.ceil(imgh / 4.0) * Math.ceil(imgw / 4.0)
let serialisedBlocks = sys.malloc((hasAlpha) ? blockCount * 20 : blockCount * 12)
let blocksWriteCount = 0
let outBlock = sys.malloc(64)
let blockSize = Math.ceil(imgh / 4.0) * Math.ceil(imgw / 4.0)
let blockWidth = Math.ceil(imgw / 4.0)
println(`Dim: ${imgw}x${imgh}, channels: ${channels}, Has alpha: ${hasAlpha}`)
// TODO write output to dedicated ptr and gzip it
let writeCount = 0
let writeBuf = sys.malloc(blockSize * ((hasAlpha) ? 20 : 12))
function chromaToFourBits(f) {
return Math.round((f * 7.5) + 7.5)
}
for (let blockY = 0; blockY < Math.ceil(imgh / 4.0); blockY++) {
for (let blockX = 0; blockX < Math.ceil(imgw / 4.0); blockx++) {
let pixelWordOffset = channels * (blockY * 4) * imgw + (blockX * 4)
let ys = Uint8Array(16)
let as = Uint8Array(16)
let cos = Float32Array(16)
let cgs = Float32Array(16)
for (let blockX = 0; blockX < Math.ceil(imgw / 4.0); blockX++) {
// println(`Encoding block ${1 + blockY * blockWidth + blockX}/${blockSize}`) // print statement is making things slower...
let ys = new Uint8Array(16)
let as = new Uint8Array(16)
let cos = new Float32Array(16)
let cgs = new Float32Array(16)
// TODO 4x4 bayer dither
for (let py = 0; py < 4; py++) { for (let px = 0; px < 4; px++) {
let offset = imageData + pixelWordOffset + 4 * (px + py * imgw)
let r = sys.peek(offset) / 255.0
let g = sys.peek(offset+1) / 255.0
let b = sys.peek(offset+2) / 255.0
let a = (hasAlpha) ? sys.peek(offset+3) / 255.0 : 1.0
// TODO oob-check
let ox = blockX * 4 + px
let oy = blockY * 4 + py
let offset = channels * (oy * imgw + ox)
let r = sys.peek(imageData + offset) / 255.0
let g = sys.peek(imageData + offset+1) / 255.0
let b = sys.peek(imageData + offset+2) / 255.0
let a = (hasAlpha) ? sys.peek(imageData + offset+3) / 255.0 : 1.0
let co = r - b // [-1..1]
let tmp = b + co / 2.0
@@ -105,17 +142,17 @@ for (let blockX = 0; blockX < Math.ceil(imgw / 4.0); blockx++) {
}}
// subsample by averaging
cos1 = Math.round((((cos[0]+cos[1]+cos[4]+cos[5]) / 4.0) + 1) * 15)
cos2 = Math.round((((cos[2]+cos[3]+cos[6]+cos[7]) / 4.0) + 1) * 15)
cos3 = Math.round((((cos[8]+cos[9]+cos[12]+cos[13]) / 4.0) + 1) * 15)
cos4 = Math.round((((cos[10]+cos[11]+cos[14]+cos[15]) / 4.0) + 1) * 15)
cgs1 = Math.round((((cgs[0]+cgs[1]+cgs[4]+cgs[5]) / 4.0) + 1) * 15)
cgs2 = Math.round((((cgs[2]+cgs[3]+cgs[6]+cgs[7]) / 4.0) + 1) * 15)
cgs3 = Math.round((((cgs[8]+cgs[9]+cgs[12]+cgs[13]) / 4.0) + 1) * 15)
cgs4 = Math.round((((cgs[10]+cgs[11]+cgs[14]+cgs[15]) / 4.0) + 1) * 15)
let cos1 = chromaToFourBits((cos[0]+cos[1]+cos[4]+cos[5]) / 8.0)
let cos2 = chromaToFourBits((cos[2]+cos[3]+cos[6]+cos[7]) / 8.0)
let cos3 = chromaToFourBits((cos[8]+cos[9]+cos[12]+cos[13]) / 8.0)
let cos4 = chromaToFourBits((cos[10]+cos[11]+cos[14]+cos[15]) / 8.0)
let cgs1 = chromaToFourBits((cgs[0]+cgs[1]+cgs[4]+cgs[5]) / 8.0)
let cgs2 = chromaToFourBits((cgs[2]+cgs[3]+cgs[6]+cgs[7]) / 8.0)
let cgs3 = chromaToFourBits((cgs[8]+cgs[9]+cgs[12]+cgs[13]) / 8.0)
let cgs4 = chromaToFourBits((cgs[10]+cgs[11]+cgs[14]+cgs[15]) / 8.0)
// append encoded blocks
let outBlock = serialisedBlocks + blocksWriteCount
// append encoded blocks to the file
let outBlock = writeBuf + writeCount
sys.poke(outBlock+ 0, (cgs1 << 4) | cos1)
sys.poke(outBlock+ 1, (ys[1] << 4) | ys[0])
@@ -139,11 +176,26 @@ for (let blockX = 0; blockX < Math.ceil(imgw / 4.0); blockx++) {
sys.poke(outBlock+17, (as[13] << 4) | as[12])
sys.poke(outBlock+18, (as[11] << 4) | as[10])
sys.poke(outBlock+19, (as[15] << 4) | as[14])
blocksWriteCount += 8
writeCount += 8
}
blocksWriteCount += 12
writeCount += 12
}}
// TODO open outfile, write header, write serialisedBlocks.gz
// write header to the output file
let headerBytes = [
0x1F, 0x54, 0x53, 0x56, 0x4D, 0x69, 0x50, 0x46, // magic
imgw & 255, (imgw >> 8) & 255, // width
imgh & 255, (imgh >> 8) & 255, // height
((hasAlpha) ? 1 : 0), 0x00, // has alpha
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00 // reserved
]
filesystem.open("A", exec_args[2], "W")
filesystem.writeBytes("A", headerBytes)
filesystem.open("A", exec_args[2], "A")
dma.ramToCom(writeBuf, 0, writeCount)
sys.free(outBlock)
sys.free(imageData)
sys.free(writeBuf)